Running gets slower, then versus now

How this is measured

For every degree above 12°C, an averagely trained recreational runner loses roughly 4.25 seconds per mile. Each capital's "then" and "now" figures apply that rate to the average hottest-month daily high across two 11-year windows, 1976 to 1986 and 2015 to 2025, using the hottest month rather than an annual average since that is when most outdoor distance races and training actually happen. Averaging over 11 years rather than picking a record year means one exceptional heatwave cannot move a city up this table on its own.

Neither number is arbitrary. 12°C is the commonly cited optimal air temperature for distance running in marathon-performance sports-science literature, and 4.25 seconds per mile per degree above it is a WBGT-based figure from studies of recreational marathon finish times against race-day temperature (e.g. Ely et al.), not a number tuned to fit this dataset.

Unlike most facts in this ledger, this one uses a fixed reference temperature and a fixed penalty rate rather than one drawn from each city's own distribution. That is a deliberate limitation worth stating: it is a population-average heuristic for a recreational runner, not a personalized forecast, and it does not model heat acclimatization, humidity or wind. Read this table as a description of how much hotter each capital's racing season has become, not as a precise pace calculator for any one runner.

Source data is the ERA5 reanalysis, retrieved through the Open-Meteo historical weather API. Only capitals with a complete daily record back to 1976 appear here; the rest are omitted rather than shown with a partial figure.